Catalytic Performance of Doped Ni₂P Surfaces for Ammonia Synthesis †

Abstract

Ammonia is a key ingredient in fertilizer production, but its synthesis using the conventional Haber–Bosch process over metal-based catalysts is energy intensive. Prior investigations revealed that metal catalysts suffer from a trade-off between N₂ activation and N* binding strength, hindering their overall reactivity. Metal phosphide catalysts are promising alternatives to conventional metal catalysts due to their unique reactivity and stability. Here, we used DFT to study the catalytic performance of Ni₂P catalysts doped with Fe and Ru for ammonia synthesis. We show that H-assisted N–N activation may provide a new route to circumvent the N₂ dissociation scaling relationships.

1. Introduction

Ammonia (NH₃) holds a crucial position in various aspects of modern society. Its significance spans from agriculture to industry, making it an essential molecule for the global economy and sustainability efforts. It serves as a fundamental building block for the production of fertilizers, which are essential for enhancing crop yields and ensuring food security for a growing global population [1]. Additionally, ammonia finds applications in diverse sectors, including chemicals, pharmaceuticals, and even renewable energy storage. The Haber–Bosch process revolutionized agricultural productivity in the early 20th century by converting atmospheric nitrogen into ammonia over Fe-based catalysts and contributing to the large-scale production of synthetic fertilizers [2]. However, despite its immense importance, ammonia synthesis presents a range of formidable challenges.

The nitrogen (N₂) dissociation step is believed to be the rate-limiting step in ammonia synthesis [3,4,5,6]. Cleaving the strong N≡N bond requires high temperatures and pressures over the conventional ammonia synthesis catalysts (Fe and Ru). The surface intermediates (NH_x; x = 0–3) formed during ammonia synthesis often bind too strongly to most transition metal catalysts under ambient conditions; thus, elevated temperatures and pressures are necessary to maintain a favorable exergonic nature of the overall reaction [7]. The catalytic reduction of dinitrogen to ammonia under mild conditions is a long-standing challenge in catalysis. The goal is to find a catalyst that can bind the NH_x species weakly while also maintaining a low energy barrier for N₂ dissociation. Computational assessments using density functional theory (DFT), however, have revealed a significant limitation: the N–N transition state energy cannot be adjusted without influencing the binding energy of NH_x to the surface due to the presence of a Brønsted–Evans–Polanyi (BEP) scaling relationship for transition metal surfaces [8,9,10,11]. Decreasing the N₂ activation barrier also increases the binding strength of NH_x, thus hindering subsequent hydrogenation reactions and the production of NH₃. These studies have concluded that the turnover rate for ammonia synthesis as a function of N* binding energy for several transition metals follows a volcano-shaped curve, with Fe and Ru being the only metals that are close to the optimum.

A logical approach to catalyst design was suggested, involving the combination of metals positioned on opposite sides of the volcano to achieve the desired interaction with nitrogen intermediates. For example, combining Co, which binds N* too weakly, with Mo, which binds N* too strongly in the Co–Mo catalyst, leads to a material that has ammonia synthesis activity similar to that of Ru [12]. Many subsequent studies have focused on finding ways to circumvent the scaling relationship using different catalytic materials, such as metal nitrides [13], hydrides [14], and oxides [15]. Electrochemical N₂ reduction has also emerged as an alternative approach for synthesizing NH₃ at ambient conditions, but the development of efficient electrochemical catalysts remains a key challenge [16,17,18]. Here, we explore the catalytic performance of the Ni₂P catalyst doped with Fe and Ru for ammonia synthesis using DFT. We have previously shown that incorporating phosphorus atoms in nickel phosphide catalysts can alter the catalytic behavior of Ni atoms, leading to improved selectivity for various catalytic reactions [19,20,21]. The P atoms introduce weak binding surface sites that disrupt the metal ensembles. We also explore H-assisted N–N activation pathways involving diazene (N₂H₂) and hydrazine (N₂H₄) to weaken the N–N bond prior to activation.

2. Computational Methods

Periodic plane-wave DFT calculations were performed using the Vienna ab initio simulation package (VASP) [22,23,24,25] and the computational catalysis interface (CCI) [26]. The RPBE form of the generalized gradient approximation and PAW pseudopotentials were used [27,28,29,30,31]. The Ni₂P(001) surface was constructed from bulk Ni₂P, as described in more detail in our previous work [19]. Figure 1 shows the top and side views of the 2 × 2 Ni₂P(001) surface used in this study. The bottom two atomic layers were constrained during geometric relaxation. Electronic energies were converged to within 10⁻⁶ eV, and forces were converged to less than 0.05 eVÅ⁻¹ with a k-point mesh of 3 × 3 × 1 [32,33]. Transition state structures were identified using the nudged elastic band (NEB) method and the dimer method [34,35,36]. The Ni₂P(001) surface was modified by substituting a surface Ni atom (highlighted in Figure 1a) with either a Fe or Ru atom. This procedure mirrors the approach employed in a prior investigation using doped Ni₂P surfaces [37]. The formation energies of doped Ni₂P surfaces were calculated using the following formula:

$$E_f=E_{\mathrm{M}/\mathrm{N}\mathrm{i}2\mathrm{P}}+E_{\mathrm{N}\mathrm{i}}-E_{\mathrm{N}\mathrm{i}2\mathrm{P}}-E_{\mathrm{M}}$$

(1)

where E_Ni and E_M are the energies of a single Ni atom and a single metal atom. The Ru–Ni₂P surface has a formation energy of −2.050 eV compared to −1.129 eV for Fe–Ni₂P, indicating that Ru–Ni₂P is more stable. Additional calculations were run using the pure 3 × 3 Fe(110) and Ru(001) surfaces for comparison. Spin-polarized calculations were exclusively applied to the Fe(110) and Fe–Ni₂P doped surfaces, as the energy discrepancy between spin-polarized and non-spin-polarized calculations exceeded 10⁻³ eV.

3. Results and Discussion

3.1. NH_x Species Binding Energies

In this section, we investigate the preferred adsorption configuration and binding energies of different NH_x species over Ni₂P, as well as doped Fe–Ni₂P and Ru–Ni₂P catalysts. For the sake of comparison, we also included the binding energies over pure Fe(110) and Ru(001) surfaces. The binding energies shown in

Table 1. The binding energies (ΔE_ads in kJ mol⁻¹) of N₂*, H*, and NH_x (x = 0–3) over the catalytic surfaces examined in this study.

	Ni2P(001)		Fe–Ni2P(001)		Ru–Ni2P(001)		Fe(110)		Ru(001)
Species	Mode	ΔEads	Mode	ΔEads	Mode	ΔEads	Mode	ΔEads	Mode	ΔEads
N2*	M1	−25	M1	−35	M1	−50	M1	−95	M1	−34
H*	M3	−222	M3	−240	M3	−246	M3	−288	M3	−262
N*	M3	−375	M3	−435	M3	−445	M4	−601	M3	−551
NH*	M3	−325	M3	−353	M3	−365	M3	−528	M3	−437
NH2*	M2	−224	M2	−235	M2	−256	M2	−326	M3	−213
NH3*	M1	−60	M1	−50	M1	−74	M1	−109	M1	−63

are calculated with respect to the gas-phase species and bare surfaces (i.e., ΔE_ads = E_species/surf − E_species(g) − E_surf). Diatomic nitrogen (N₂) vertically adsorbs on the metal atop site (M₁) over Ni₂P with a binding energy of −25 kJ mol⁻¹. This binding energy increases to −35 kJ mol⁻¹ for Fe–Ni₂P and −50 kJ mol⁻¹ for Ru–Ni₂P. For the NH_x* species, the binding strength increases as the number of hydrogen atoms bonded to the N atom decreases, i.e., NH₃* binds most weakly to the surface, whereas N* has the highest binding energy. This reflects the increasing stability of the more saturated NH_x* species in the gas phase, and they coordinate with fewer metal atoms upon adsorption. For example, N* and NH* bind to the metal 3-fold (M₃) site, whereas NH₂ and NH₃ bind to the bridging (M₂) and atop (M₁) sites, respectively (Figure 2).

We also observe a general trend in binding energies across examined surfaces where the binding strength for each species generally increases as Ni₂P → Fe–Ni₂P → Ru–Ni₂P → Ru → Fe with few exceptions. If we consider the N* binding energy as a proxy for the catalytic activity towards N–N bond activation within N₂, following the principles of Brønsted–Evans–Polanyi (BEP) scaling relations [8,9,10,11], it can be inferred that Ni₂P is likely to exhibit weak activity. Conversely, Fe and Ru-doped Ni₂P catalysts are expected to demonstrate higher activity compared to pure Ni₂P, although falling short of the reactivity of Fe- and Ru-based catalysts. However, this also means that Fe- and Ru-based catalysts are more prone to deactivation due to the strongly bound N* or NH* than phosphide-based catalysts.

3.2. N–N Activation Reactions

Here, we examine the N–N bond cleavage reactions in N₂*, N₂H₂*, and N₂H₄*, as shown in Figure 3. Over the Ru(001) surface, adsorbed N₂* undergoes N–N bond activation with a forward activation barrier (ΔE_a) of 168 kJ mol⁻¹ to form two nitrogen atoms (2N*) both adsorbed on separate 3-fold sites. This reaction is slightly endothermic with a reaction energy (ΔE_rxn) of 13 kJ mol⁻¹. These results are consistent with previous studies on Ru(001) flat surfaces [38]. On both Fe-Ni₂P and Ru-Ni₂P surfaces, N₂* dissociates to 2N* with a much larger activation barrier of 350 kJ mol⁻¹ and a highly endothermic reaction energy. This reaction starts with N₂* vertically adsorbed on the doped metal atom. After the N–N bond cleavage, the N* atoms bind to nearby 3-fold sites. We have attempted this reaction and even the other H-assisted N–N bond activation reactions in N₂H₂ and N₂H₄ over the undoped Ni₂P surface, but we were not able to identify stable transition states. The trends in N* binding energy discussed in Section 3.1 indeed correlate with the calculated N–N bond activation barriers over these materials.

We next examine the N–N bond activation reactions in diazine (N₂H₂) and hydrazine (N₂H₄) to investigate whether hydrogenating N₂ can weaken the N–N bond prior to its activation. Figure 3 shows that the activation barrier of the N–N bond in adsorbed N₂H₂* dramatically decreased by more than 220 kJ mol⁻¹ over both Fe–Ni₂P and Ru–Ni₂P surfaces (121 and 129 kJ mol⁻¹, respectively) compared to direct N₂* dissociation (350 kJ mol⁻¹). Breaking the N–N bond in N₂H₄ also exhibits a similar barrier of 128 kJ mol⁻¹ over Ru-Ni₂P, but the barrier for this reaction over Fe-Ni₂P increases to 143 kJ mol⁻¹. We expect a similar decrease in the activation barrier of H-assisted N–N activation over a pure Ru(001) surface; however, the resulting NH* also strongly binds to the surface (Table 1), which may render the active sites inert. The more facile activation of the N–N bond in N₂H₂* over both Fe–Ni₂P and Ru-Ni₂P-doped surfaces, coupled with the weaker binding of the product NH*, may open the door for a promising pathway that avoids the current paradigm of direct N₂ dissociation. However, it is important to note that further analysis of the hydrogenation reactions to produce N₂H₂ and eventually NH₃ over these materials is necessary to fully ascertain the viability of this pathway.

4. Conclusions

In this work, we investigated the binding energies of various NH_x species and the N–N bond activation energy over Ni₂P catalysts doped with Fe and Ru using DFT. Our results indicate that the NH_x species with fewer hydrogen atoms bind more strongly to the catalytic surfaces, and the binding strength generally increases as Ni₂P → Fe–Ni₂P → Ru–Ni₂P → Ru → Fe. This trend is reflected in the relative activity of these catalytic surfaces towards direct N₂* dissociation, with Ru being the most active and Ni₂P the least active surface. However, we show that hydrogenating the N₂* prior to N–N bond cleavage can dramatically decrease the activation barrier over doped Ni₂P surfaces. These findings suggest that the H-assisted N–N bond activation may play a crucial role in reducing the activation energy required for ammonia synthesis over these surfaces without a concomitant increase in the binding strength of the NH_x species. Further investigation into the underlying mechanisms of this catalytic process could potentially lead to enhanced strategies for nitrogen activation and ammonia synthesis.